CN110888398B

CN110888398B - Control device, CNC device and control method

Info

Publication number: CN110888398B
Application number: CN201910842338.4A
Authority: CN
Inventors: 於保勇作; 恒木亮太郎; 奥田真司
Original assignee: Fanuc Corp
Current assignee: Fanuc Corp
Priority date: 2018-09-11
Filing date: 2019-09-06
Publication date: 2024-04-30
Anticipated expiration: 2039-09-06
Also published as: CN110888398A; DE102019213431A1; US11287801B2; US20200081411A1; JP2020042585A; JP6856591B2

Abstract

The invention provides a control device, a CNC device and a control method. The control device is provided with: a machine learning unit that performs machine learning of control parameters that determine operation characteristics of a drive unit of a machine that is a drive target of a motor, and sets the control parameters to a motor control device that controls the motor; an operation state check operation instruction unit that outputs an instruction to drive the operation state check operation of the motor control device; an operation evaluation unit that acquires information indicating operation characteristics from a motor control device or a machine that is driven in response to an instruction of an operation condition inspection operation, calculates an evaluation value from an evaluation function using the information, and stores the evaluation value in a storage unit in association with time information at the time of executing the operation condition inspection operation or operation information of the motor; and a degradation estimating unit that estimates degradation of the operation characteristic based on the evaluation value calculated by the operation evaluating unit and the evaluation value stored in the storage unit when the operation condition checking operation is performed.

Description

Control device, CNC device and control method

Technical Field

The present invention relates to a control device for performing an operation condition check operation of a motor control device for driving and controlling a motor in order to estimate deterioration of an operation characteristic of a driving unit of a machine to be driven by the motor, a CNC (computer numerical control) device using the control device, and a control method of the control device.

Background

For example, patent documents 1 to 3 describe devices for detecting faults or abnormalities in machines such as machine tools and industrial machines.

Patent document 1 describes a machine learning device capable of detecting a sign of a failure before the failure of a spindle of a machine tool or a motor driving the spindle. Specifically, a machine learning device that learns a failure prediction of a spindle or a motor that drives the spindle of a machine tool includes: a state observation unit configured to observe a state variable including at least one of output data of a motor control device that controls the motor, output data of a detector that detects a state of the motor, and output data of a measurer that measures a state of the spindle or the motor; a determination data acquisition unit that acquires determination data obtained by determining the presence or absence of a failure or the degree of the failure of the spindle or the motor; and a learning unit for learning the failure prediction of the spindle or motor based on the data set created from the combination of the state variables and the determination data.

Patent document 2 describes a failure prediction system capable of accurately predicting a failure according to a situation. Specifically, the failure prediction system includes a machine learning device that learns conditions associated with a failure of the industrial machine. The machine learning device is provided with: a state observation unit configured to observe a state variable including output data of a sensor, internal data of control software, or calculation data obtained from the output data of the sensor, the internal data, or the internal data, during operation or rest of the industrial machine; a determination data acquisition unit that acquires determination data indicating the presence or absence of a fault or the degree of the fault in the industrial machine; and a learning unit that learns conditions associated with a failure of the industrial machine by supervised learning, in accordance with a training data set created from a combination of the state variables and the determination data.

Patent document 3 describes an abnormality detection device for a tool of a machine tool, which can improve the diagnostic accuracy. Specifically, the abnormality detection device includes: an acquisition unit that acquires a plurality of measured values related to a tool as measured data (vibration information, cutting force information, sound information, spindle load, motor current, power value, etc.); a normal model unit that learns measurement data acquired during normal processing by class 1 machine learning, and creates a normal model; an abnormality diagnosis unit that obtains measurement data at the time of machining after the production of the normal model, and diagnoses whether the measurement data is normal or abnormal based on the normal model; and a re-diagnosis unit that re-diagnoses the measurement data diagnosed as abnormal by the abnormality diagnosis unit by a method different from the abnormality diagnosis unit.

Patent documents 1 to 3 describe apparatuses for detecting a failure or abnormality of a machine tool or an industrial machine by machine learning, but since the amount of information processing is large, machine learning requires a certain learning time, and the operation efficiency of the machine is lowered in order to perform machine learning.

Patent document 1: japanese patent laid-open No. 2017-188030

Patent document 2: japanese patent laid-open No. 2017-120649

Patent document 3: japanese patent application laid-open No. 2018-24055

Disclosure of Invention

The invention aims at: provided are a control device capable of estimating deterioration of the operation characteristics of a drive unit of a machine such as a machine tool or an industrial machine without performing machine learning, a CNC device using the control device, and a control method of the control device.

(1) The control device (for example, a control device 500 described below) of the present invention includes:

A machine learning unit (for example, a machine learning unit 550 described later) that performs machine learning on control parameters that determine the operation characteristics of a driving unit of a machine (for example, a machine 400 described later) that is a driving target of a motor (for example, a servo motor 300 described later), and sets the control parameters to a motor control device (for example, a servo control device 200 described later) that controls the motor;

an operation state check operation instruction unit (for example, an operation state check operation instruction unit 510 described later) that outputs an instruction to drive an operation state check operation of the motor control device in order to obtain an operation characteristic of the drive unit;

An operation evaluation unit (for example, an operation evaluation unit 520 described later) that obtains information indicating an operation characteristic of the driving unit from the motor control device or the machine that is driven in response to an instruction of the operation condition checking operation, calculates an evaluation value from an evaluation function using the information, and stores the evaluation value in a storage unit (for example, a storage unit 530 described later) in association with time information at which the operation condition checking operation is performed or operation information of the motor;

And a degradation estimating unit (for example, degradation estimating unit 540 described later) that estimates degradation of the operation characteristics of the driving unit of the machine based on the evaluation value calculated by the operation evaluating unit at the time of the operation condition checking operation and the evaluation value stored in the storage unit.

(2) In the control device according to (1), the degradation estimating unit may instruct the machine learning unit to learn the control parameter based on a result of estimating degradation of the operation characteristic of the driving unit of the machine.

(3) In the control device according to the above (1) or (2), the degradation estimating unit may instruct a report unit (for example, a report unit 560 described later) for notifying degradation of the operation characteristic of the driving unit of the machine, based on a result of estimating the degradation of the operation characteristic of the driving unit of the machine.

(4) In the control device according to any one of (1) to (3), the operation state check operation instruction unit may output an instruction of the operation state check operation when a predetermined signal is input or according to a predetermined schedule.

(5) In the control device according to any one of (1) to (4), the operation state check operation instruction unit may transmit an instruction of the operation state check operation to a numerical control device that outputs a control instruction to the motor control device.

(6) The CNC device of the present invention comprises: the control device of the above (5); a motor control device that controls the motor; and a numerical controller for outputting a control command to the motor controller based on the instruction of the operation condition check operation outputted from the controller.

(7) The control method of the present invention is performed by a control device including a storage unit (for example, a storage unit 530 described later) and a machine learning unit (for example, a machine learning unit 550 described later) for performing machine learning on control parameters that determine operation characteristics of a driving unit of a machine (for example, a machine 400 described later) to be driven by a motor (for example, a servo motor 300 described later) and setting the control parameters to a motor control device that controls the motor, the control method including:

An operation state check operation instruction step of outputting an instruction to drive an operation state check operation of the motor control device in order to obtain an operation characteristic of the driving unit;

an operation evaluation step of acquiring information indicating an operation characteristic of the driving unit from the motor control device or the machine driven in accordance with the instruction of the operation condition checking operation, calculating an evaluation value based on an evaluation function using the information, and storing the evaluation value in a storage unit in association with time information at the time of executing the operation condition checking operation or the operation information of the motor;

And a degradation estimation step of estimating degradation of the operation characteristic of the driving unit of the machine based on the evaluation value calculated in the operation evaluation step and the evaluation value stored in the storage unit when the operation condition checking operation is performed.

(8) In the control method of the control device according to (7), the degradation estimation step may further include: and a machine learning instruction step of machine learning for instructing the machine learning unit to perform the control parameter based on the result of the estimation of the deterioration of the operation characteristic of the driving unit of the machine estimated in the deterioration estimation step.

(9) In the control method of the control device according to (7), the degradation estimation step may further include: and a reporting step of notifying deterioration of the operation characteristic of the driving unit of the machine based on a result of estimating the deterioration of the operation characteristic of the driving unit of the machine in the deterioration estimating step.

According to the present invention, deterioration of the operation characteristics of the drive unit of a machine such as a machine tool or an industrial machine can be estimated without performing machine learning.

Drawings

Fig. 1 is a block diagram showing a machine control system including a machine control device according to a first embodiment of the present invention.

Fig. 2 is a block diagram showing the configuration of a numerical controller, a servo controller, and a machine in the machine control system.

Fig. 3 is a diagram for explaining a movement locus of a square shape with rounded corners.

Fig. 4 is a diagram showing a square movement locus with rounded corners in the case where adjustment by machine learning of coefficients ω, τ, and σ of a transfer function of a filter of a servo control device and thereafter aging degradation of a machine occur.

Fig. 5 is a characteristic diagram showing an example of the positional deviation e used in the evaluation function of formula 3.

Fig. 6 is a characteristic diagram showing an example of the positional deviation e used in the evaluation function of equation 4.

Fig. 7 is a characteristic diagram showing a state in which an evaluation value at the time of machine learning before shipment rises due to aged deterioration, and a state in which an evaluation value at the time of relearning rises again due to aged deterioration.

Fig. 8 is a flowchart showing the operation of the control device according to the present embodiment.

Fig. 9 is a diagram for explaining a circular movement locus.

Fig. 10 is a diagram for explaining a square movement locus.

Fig. 11 is a diagram for explaining an octagonal movement path.

Fig. 12 is a diagram for explaining a movement locus of a star.

Fig. 13 is a block diagram showing a machine learning unit according to the first embodiment of the present invention.

Fig. 14 is a block diagram showing the configuration of a numerical controller, a servo controller, and a machine in a machine control system according to a second embodiment of the present invention.

Fig. 15 is a block diagram showing another configuration example of the machine control system.

Description of the reference numerals

10. 10A: a mechanical control system; 20. 20-1 to 20-n: CNC means; 100: a numerical control device; 200: a servo control device; 300: a servo motor; 400. 400-1 to 400-n: a machine; 500. 500-1 to 500-n: a control device; 510: an operation status checking operation instruction unit; 520: an operation evaluation unit; 530: a storage unit; 540: a degradation estimation unit; 550: a machine learning unit; 560: and a reporting unit.

Detailed Description

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

(First embodiment)

Fig. 1 is a block diagram showing a machine control system including a machine control device according to a first embodiment of the present invention, and fig. 2 is a block diagram showing the configuration of a numerical controller, a servo controller, and a machine in the machine control system.

As shown in fig. 1, the machine control system 10 includes a numerical controller 100, a servo controller 200, a servo motor 300, a machine 400, and a controller 500. The servo control device 200 is a motor control device. The numerical controller 100, the servo controller 200, and the servo motor 300 constitute the CNC device 20. Further, the control device 500 may be included in the CNC device 20. The machine 400 is a driving target of the servomotor 300.

First, before explaining the control device 500, the numerical control device 100, the servo control device 200, and the machine 400 will be explained.

The numerical controller 100 includes a numerical control processing unit 101 and a storage unit 102. The storage unit 102 stores a machining program for checking the operation state, a machining program for machine learning, and a machining program in the case of actually performing machining processing. The numerical control processing unit 101 receives an instruction of the operation state check operation from the operation state check operation instruction unit 510 of the control device 500, reads a machining program for operation state check from the storage unit 102, generates a position instruction to be a control instruction, and outputs the position instruction to the servo control device 200.

The numerical control processing unit 101 receives a machine learning instruction from the machine learning unit 550 of the control device 500 or a machining instruction from an operator at the time of machine learning or actual machining, reads a machining program for machine learning or actual machining from the storage unit 102, generates a position instruction, and outputs the position instruction to the servo control device 200. Further, the numerical control processing unit 101 may receive the machine learning execution instruction from the degradation estimation unit 540 instead of the machine learning unit 550.

As shown in fig. 2, the servo control device 200 includes: a filter 201, a subtractor 202, a position control unit 203, an adder 204, a differentiator 205, a position feedforward processing unit 206, a subtractor 207, a speed control unit 208, and an integrator 209. The subtractor 202, the position control unit 203, the adder 204, the subtractor 207, the speed control unit 208, the servo motor 300, the rotary encoder 310, and the integrator 209 constitute a position feedback loop. The subtractor 207, the speed control unit 208, the servo motor 300, and the rotary encoder 310 form a speed feedback loop.

The filter 210 is a vibration damping filter that suppresses vibration of the machine 400, and is a band elimination filter shown by a mathematical formula (1) (hereinafter, expressed as formula 1), for example.

In mathematical formula 1, coefficient s is a coefficient of time, which is an argument of an exponential function in the laplace transform, coefficient ω is a mask center angular frequency, coefficient τ is a bandwidth ratio, and coefficient g is g=στ. If the mask center frequency is fc and the mask frequency bandwidth is fw, the coefficient ω is represented by ω=2pi fc and the coefficient τ is represented by τ=fw/fc. The coefficient sigma is a vibration reduction coefficient (0 < sigma.ltoreq.1). Coefficients ω, τ, and σ of the transfer function of the filter 201 are control parameters for determining the operation characteristics of the driving unit of the machine 400 by a servo control device serving as a motor control device.

[ Formula 1]

A position command is input from the numerical control processing unit 101 to the filter 201. Here, the filter 201 is disposed outside the position feedback loop and the speed feedback loop, but may be disposed in the position feedback loop or the speed feedback loop. For example, the filter 201 may be connected to an output side of a speed control unit 208 or an output side of the adder 204, which will be described later. The configuration of the filter 201 is not particularly limited, but an IIR filter of second order or more is preferable.

The subtractor 202 obtains a difference between the shaped position command output from the filter 201 and the position feedback detection position, and outputs the difference as a positional deviation to the position control unit 203, a machine learning unit 550, and an operation evaluation unit 520, which will be described later. The operation evaluation unit 520 performs an operation condition check operation using the positional deviation, and the machine learning unit 550 performs machine learning using the positional deviation.

The position control unit 203 outputs a value obtained by multiplying the position deviation by the position gain Kp to the adder 204 as a speed command value.

The differentiator 205 outputs a value obtained by differentiating the position command value and multiplying the value by the constant α to the position feedforward processing section 206. The position feedforward processing unit 206 performs a position feedforward process shown by a transfer function G(s) shown by a mathematical formula (2) (hereinafter, expressed as formula 2) on the output of the differentiator 205, and outputs the result of the process to the adder 204 as a position feedforward term. The coefficients a _i、b_j (X.gtoreq.i, j.gtoreq.0, X is a natural number) of equation 2 are coefficients of the transfer function G(s).

[ Formula 2]

The adder 204 adds the speed command value and the output value (position feedforward term) of the position feedforward processing unit 206, and outputs the result to the subtractor 207 as a feedforward-controlled speed command value. The subtractor 207 obtains a difference between the output of the adder 204 and the speed detection value of the speed feedback, and outputs the difference to the speed control unit 208 as a speed deviation.

The speed control unit 208 multiplies the speed deviation by an integral gain K1v, and outputs a value obtained by integrating the speed deviation and the integral gain K2v to the servo motor 300 as a torque command.

The rotational angle position of the servomotor 300 is detected by the rotary encoder 310, and the speed detection value is input as a speed feedback (speed FB) to the subtractor 207. The speed detection value is integrated by the integrator 209 to become a position detection value, and the position detection value is input to the subtractor 202 as position feedback (position FB).

As described above, the servo control device 200 is configured.

The servo motor 300 is rotationally controlled by the servo controller 200, and drives the machine 400.

The machine 400 is a machine tool, an industrial machine, or the like. The servo motor 300 may be incorporated into a machine 400 such as a machine tool or an industrial machine. In fig. 2, a machine 400 is represented as a machine tool, a portion of which is shown. As shown in fig. 2, the machine 400 includes a ball screw 401 coupled to a rotation shaft of the servomotor 300, a nut 402 screwed to the ball screw 401, and a table 403 connected to the nut 402. The ball screw 401, the nut 402, and the table 403 constitute a driving section. By the rotational drive of the servomotor 300, the nut 402 screwed to the ball screw 401 moves in the axial direction of the ball screw 401.

In the machine 400, when a table 403 on which a workpiece (work) is mounted is moved in the X-axis direction and the Y-axis direction, the servo control device 200 and the servo motor 300 shown in fig. 2 are provided for the X-axis direction and the Y-axis direction, respectively. When the table is moved in a direction of 3 axes or more, the servo control device 200 and the servo motor 300 are provided for each axis direction.

Next, the control device 500 is described.

First, the purpose of the operation state checking operation of the control device 500 will be described. Here, the operation condition checking operation is an operation of driving the servo control device using a machining program for operation condition checking, acquiring information indicating an operation characteristic of a driving portion of the machine, obtaining an evaluation value from an evaluation function using the information, and estimating the aged deterioration of the machine from the evaluation value. Further, as an example of deterioration of the operation characteristics, aging deterioration due to use of a machine is exemplified, but the present invention is not limited thereto, and deterioration of the operation characteristics due to changes in the surrounding environment such as temperature, humidity, vibration, and the like is also included.

The machining program for checking the operation state stored in the storage unit 102 of the numerical controller 100 is a machining program in which the servo controller 200 controls the servo motor 300 so that the movement locus ML of a specific point on the table 403 becomes a square (a square with quarter arc) with rounded corners as shown in fig. 3. In the movement path ML shown in fig. 3, the servo motor for moving the table in the Y-axis direction is decelerated when it exceeds the position P1 at the position P1, and the rotation is stopped at the position P2. On the other hand, at the position P1, the servomotor that moves the table in the X-axis direction starts rotating, and the rotation speed becomes constant at the position P2. The machining program for operation condition inspection does not load the work onto the table 403 and does not perform machining of the work.

Fig. 4 is a diagram showing a square movement locus with rounded corners in the case where the adjustment based on machine learning of coefficients ω, τ, and σ of the transfer function of the filter 201 of the servo control device 200 and the aging degradation of the machine 400 thereafter occur. As shown in fig. 4, if the servo control device 200 is operated by using a machining program for operation condition inspection in a state where the coefficients ω, τ, and σ of the transfer function of the filter 201 of the servo control device 200 are not adjusted, a protrusion called a quadrant protrusion is generated in a square rounded portion in the movement locus ML1 of a specific point on the table 403. Before shipment of the machine 400, the servo control device 200 is operated using a program for machine learning including a machining program for operation condition inspection so that no quadrant protrusion is generated in the movement locus, and a machine learning unit 550 of the control device 500 performs machine learning on optimal values of coefficients ω, τ, and σ of a transfer function of a filter 201 of the servo control device 200 so that the movement locus ML2 without quadrant protrusion is obtained by adjusting the coefficients ω, τ, and σ of the transfer function to the optimal values. However, due to aging degradation caused by the use of the machine 400, when coefficients ω, τ, and σ of the transfer function of the filter 201 set at the time of shipment are held, vibration or the like is generated in the machine, and quadrant protrusion is generated as in the movement trace ML 3.

Therefore, the control device 500 estimates the aging degradation of the machine by checking the operation state. Then, the coefficients ω, τ, and σ of the transfer function of the filter 201 are adjusted by machine learning again (referred to as relearning) based on the estimation result of the aged deterioration.

The control device 500 may perform processing such as reporting to an Operator or a manager, in place of or together with adjustment of the coefficients of the filter 201 based on machine learning, based on the estimation result of the aged deterioration.

Further, since machine learning requires a certain learning time due to a large amount of information processing, if re-learning is frequently performed, the operation efficiency of the machine is lowered for re-learning. In addition, if the report is frequently made, the load of an operator or manager as an operator monitoring the report increases.

Accordingly, the control device 500 performs the operation condition checking operation so that relearning or reporting is performed as necessary.

Next, the structure and operation of the control device 500 will be described.

As shown in fig. 1, the control device 500 includes an operation state check operation instructing unit 510, an operation evaluating unit 520, a storage unit 530, a degradation estimating unit 540, a machine learning unit 550, and a reporting unit 560.

The operation state check operation instructing unit 510 transmits an instruction of an operation state check operation to the numerical control processing unit 101 of the numerical control device 100 at the time of starting the machine 400 or in response to an instruction of execution from an operator. When the operation is started, a start signal is input to the operation status check operation instructing unit 510, or when execution is instructed from the operator, an execution instruction signal is input to the operation status check operation instructing unit 510. The start signal and the execution instruction signal correspond to predetermined signals. The operation state check operation instructing unit 510 may instruct the numerical control processing unit 101 to perform the operation state check operation according to a predetermined schedule set by the operator.

When receiving the instruction of the operation state checking operation, the numerical control processing unit 101 reads out the machining program for operation state checking from the storage unit 102, generates a position instruction, and outputs the position instruction to the servo control device 200. The servo control device 200 controls rotation of the servo motor 300 according to the position command.

The operation evaluation unit 520 obtains the positional deviation E, which is the output of the subtractor 202 of the servo control device 200, and obtains an evaluation value E _A using an evaluation function.

The evaluation function may be, for example, the evaluation function of equation 3 or equation 4 shown below. The coefficients Ca, cb and Cc in the formula 3 and Cd in the formula 4 are weighting coefficients, respectively. The coefficient t ₀ in equation 4 is a time from the change of the speed command value of the servo motor 300 to the convergence of the positional deviation e to a predetermined range. The evaluation function is not limited to the evaluation function shown in equation 3 or equation 4. For example, the time integral of the absolute value of the position deviation e of the formulas 3 and 4 may use the time integral of the square of the absolute value of the position deviation e, or the maximum value of the set of the absolute values of the position deviation e.

[ Formula 3]

Ca×∫|e|dt+Cb×∫|de/dt|dt

[ Equation 4]

Cc×∫|e|dt+Cd×∫t₀dt

Fig. 5 is a characteristic diagram showing an example of the positional deviation e used in the evaluation function of formula 3. Fig. 6 is a characteristic diagram showing an example of the positional deviation e used in the evaluation function of equation 4.

In the evaluation function of equation 3, the positional deviation shown by the broken line in fig. 5 is less vibrated than the positional deviation shown by the solid line, and the value of the time integral of i de/dt i is small and the evaluation value is small. Therefore, it was evaluated as a better result. In the evaluation function of equation 4, the positional deviation shown by the broken line in fig. 6 is larger in instantaneous deviation than the positional deviation shown by the solid line, but the static deviation is smaller, the time integral value of t ₀ is smaller, and the evaluation value is smaller. Therefore, it was evaluated as a better result.

The evaluation value E _A obtained from the evaluation function of the formula 3 or the formula 4 is output to the degradation estimation unit 540 and the storage unit 530.

The operation evaluation unit 520 causes the storage unit 530 to store, in association with each other, the evaluation value E _B based on the positional deviation E obtained by operating the servo control device 200 using the machining program for operation condition inspection after machine learning before shipment, and time information at the time of execution of the operation condition inspection operation, or operation information of the servo motor 300. The operation evaluation unit 520 causes the storage unit 530 to store, after shipment, the evaluation value E _A based on the positional deviation E obtained by operating the servo control device 200 each time using the machining program for operation condition inspection, and time information at the time of executing the operation condition inspection operation, or operation information of the servo motor 300, respectively. After the relearning, the operation evaluation unit 520 also stores the evaluation value E _A based on the positional deviation E obtained by operating the servo control device 200 each time using the machining program for operation condition inspection, and the time information at the time of executing the operation condition inspection operation, or the operation information of the servo motor 300, respectively.

The degradation estimation unit 540 obtains the evaluation value E _A and the time information at the time of executing the operation condition checking operation calculated by the operation evaluation unit 520, and the time information and the evaluation value E _B at the time of executing the operation condition checking operation after machine learning before shipment, and the time information and the evaluation value E _A at the time of executing the operation condition checking operation after shipment stored in the storage unit 530, and creates degradation estimation characteristics (degradation estimation characteristic lines) as shown in fig. 7 from these evaluation values E _B and E _A and the time information, for example, by linear approximation. The degradation estimation unit 540 similarly creates a degradation estimation characteristic (degradation estimation characteristic line) after the re-learning. Fig. 7 is a characteristic diagram showing a state in which an evaluation value after machine learning before shipment rises due to aged deterioration, and a state in which an evaluation value at the time of relearning rises again due to aged deterioration.

In the case of obtaining the degradation estimation characteristic, an n-degree curve may be assumed, and the coefficient thereof may be determined by using a least square method, thereby obtaining the degradation estimation characteristic. Further, instead of the time information when the operation condition checking operation is performed, operation information such as the movement amount of the servomotor 300 may be used.

The degradation estimation unit 540 estimates degradation of the operation characteristics of the driving unit of the machine 400 based on the degradation estimation characteristic line or the degradation estimation characteristic curve, and instructs the machine learning unit 550 to learn the control parameters again based on the estimation result. Specifically, as shown in fig. 7, the degradation estimation unit 540 estimates a learning timing based on, for example, a degradation estimation characteristic line, and sends a machine learning instruction to the machine learning unit 550. Specifically, the machine learning unit 550 is configured to send a machine learning instruction (relearning instruction) before the set degradation determination level is reached (for example, several weeks before the degradation determination level is reached) based on the degradation estimation characteristic line of the evaluation value obtained by the operation condition inspection operation before shipment. The machine learning unit 550 relearns at a timing when the machine 400 is not performing machining. The degradation determination level can be appropriately set for the processing condition of the workpiece, the transition of the evaluation value, and the like.

After relearning, the degradation estimation unit 540 similarly obtains a degradation estimation characteristic line, estimates the next learning period from the degradation estimation characteristic line, and sends a machine learning instruction to the machine learning unit 550.

Instead of or in addition to transmitting the machine learning instruction, the degradation estimation unit 540 may transmit the report instruction to the report unit 560.

The degradation estimation unit 540 may transmit the machine learning instruction to the numerical control processing unit 101 of the numerical control device 100 simultaneously with the transmission of the machine learning instruction to the machine learning unit 550. In this case, the machine learning unit 550 may not transmit the machine learning execution instruction to the numerical control processing unit 101.

When the report section 560 receives the report instruction, it displays the report instruction to the liquid crystal display device or transmits a notification of inspection, a warning of deterioration, or the like to the mobile terminal via the communication section, thereby reporting the report instruction to the operator or the manager. In the case where the machine learning unit 550 performs relearning without reporting to the operator or the manager, the reporting unit 560 may be omitted. The Operator (Operator) who has received the report may give a machine learning instruction to the machine learning unit 550 to learn again at a timing when the machine is not performing machining.

When receiving the machine learning instruction, the machine learning unit 550 transmits a machine learning execution instruction to the numerical control processing unit 101 of the numerical control device 100. When receiving the machine learning execution instruction, the numerical control processing unit 101 reads out the machining program for machine learning from the storage unit 102, generates a position instruction, and outputs the position instruction to the servo control device 200. The servo control device 200 controls rotation of the servo motor 300 according to the position command.

The machine learning unit 550 obtains the positional deviation e, which is the output of the subtractor 202 of the servo control device 200, performs machine learning on the optimal values of the coefficients ω, τ, and σ of the transfer function of the filter 201 of the servo control device 200, and adjusts the coefficients ω, τ, and σ of the transfer function to the optimal values. The machine learning will be described in detail later.

Fig. 8 is a flowchart showing the operation status checking operation of the control device 500 according to the present embodiment.

First, in step S11, the operation evaluation unit 520 obtains the positional deviation E, which is the output of the subtractor 202 of the servo control device 200 driven in response to the instruction of the operation condition check operation, and obtains the evaluation value E _A using the evaluation function.

Next, in step S12, the degradation estimation unit 540 creates degradation estimation characteristics (degradation estimation characteristic lines) using the evaluation values after machine learning before shipment. In step S13, the degradation estimation unit 540 estimates a learning period from the degradation estimation characteristic line, and transmits a machine learning instruction to the machine learning unit 550.

In step S14, the machine learning unit 550 performs machine learning based on the machine learning instruction.

In step S15, the control device 500 determines whether to continue the estimation of the aging degradation based on the instruction of the operator or the like, and if the estimation is continued (yes in step S15), the process returns to step S11, and the operation evaluation unit 520 waits until the positional deviation e is obtained based on the instruction of the next operation condition check operation.

If the estimation of the aged deterioration is not continued (no in step S15), the operation condition checking operation of the control device 500 is ended.

Next, the machine learning unit 550 of the control device 500 will be described.

< Machine learning section 550>

The machine learning unit 550 executes a machining program for machine learning, and performs machine learning (hereinafter referred to as learning) on the coefficients ω, τ, and σ of the transfer function of the filter 201 using the positional deviation obtained from the subtractor 202, so that the positional deviation becomes small. The machine learning unit 550 is a machine learning device. The learning of the machine learning unit 550 is performed before shipment, but is performed again after shipment.

Here, the movement trajectory specified by the machining program at the time of learning is, for example, circular, square, octagon, star, or the like shown in fig. 9 to 12, in addition to square with rounded corners shown in fig. 3. The learning is performed by appropriately combining the processing programs at the time of learning shown in fig. 3 and 9 to 12.

Based on the movement trajectory specified by the machining program at the time of learning, it is possible to evaluate vibration generated when the rotation direction of the servomotor for moving the table in the X-axis direction and/or the Y-axis direction is reversed or stopped from the rotation state, and to investigate the influence on the positional deviation.

By executing the machining program at the time of learning shown in fig. 3 and 9 to 12 at the time of learning, the numerical control information processing section 1011 sequentially outputs, for example, position command values so as to become a movement locus of a square shape with rounded corners, a circle shape, a square shape, an octagon shape, and a star shape.

The machine learning unit 550 performs Q learning (Q-learning) in which the state S is a servo state including values of the coefficients ω, τ, and σ of the transfer function of the filter 201, and instruction and feedback of positional deviation information of the servo control device 200 obtained by executing a machining program at the time of learning, and the like, and the behavior a is an adjustment of the coefficients ω, τ, and σ in the state S. As is well known to those skilled in the art, in Q learning, the purpose is to: in a certain state S, the behavior a having the highest value Q (S, a) is selected from among the behaviors a that can be taken as the optimal behavior.

Specifically, the engine (machine learning device) selects various behaviors a in a certain state S, and selects a better behavior based on the return given to the behavior a at that time, thereby learning the correct value Q (S, a).

Further, it is desirable to maximize the total of the returns obtained in the future, and thus the target is to finally be Q (S, a) =e Σ (γ ^t)r_t): where E [ ] represents an expected value, t is time, γ is a parameter called discount rate, which will be described later, r _t is the return at time t, and Σ is the total at time t.

[ Equation 5]

In the above formula 2, S _t represents the state of the environment at time t, and a _t represents the behavior at time t. With behavior a _t, a state change to S _t+1.r_t+1 represents a return due to the change in state. The term with max is the result of multiplying the Q value by γ when the behavior a having the highest Q value known at this time is selected in the state S _t+1. Here, γ is a parameter of 0< γ+.ltoreq.1, called discount rate. In addition, α is a learning coefficient, and is in the range of 0< α.ltoreq.1.

Equation 2 above represents a method in which the return r _t+1 returned based on the results of action a _t updates the value Q of action a _t in state S _t (S _t,A_t).

The machine learning unit 550 observes state information S including a servo state such as a feedback and a command including positional deviation information of the servo control device 200 obtained by executing a machining program at the time of learning, based on coefficients ω, τ, and σ of the transfer function of the filter 201, and determines the behavior a. The status information corresponds to the feedback information. The machine learning unit 550 returns the return r each time the action a is performed.

Here, the return r is set as follows.

When the state information S is corrected to the state information S 'based on the behavior information a, the value of the evaluation function of the position deviation e of the servo control device 200 operated based on the corrected coefficients ω, τ, and σ related to the state information S' is set to be a negative value when the value of the evaluation function of the position deviation e of the servo control device 200 operated based on the corrected coefficients ω, τ, and σ related to the state information S before correction based on the behavior information a is larger than the value of the evaluation function of the position deviation e of the servo control device 200 operated based on the corrected coefficients ω, τ, and σ related to the state information S before correction based on the behavior information a. The evaluation function may be, for example, the evaluation function of equation 3 or equation 4 used in the operation evaluation unit 520. However, an evaluation function different from the evaluation function used in the operation evaluation unit 520 may be used for the evaluation function used in the machine learning.

On the other hand, when the value of the evaluation function of the position deviation e of the servo control device 200 operating based on the corrected coefficients ω, τ, and σ related to the state information S' corrected based on the behavior information a is smaller than the value of the evaluation function of the position deviation e of the servo control device 200 operating based on the coefficients ω, τ, and σ related to the state information S before correction based on the behavior information a, the value of the return is set to a positive value.

In Q learning, the machine learning unit 505 searches for the optimal behavior a with the largest total of the future returns r, for example, by trial and error. Thus, the machine learning unit 505 can select the optimal behavior a (i.e., the values of the coefficients ω, τ, and σ of the transfer function of the optimal filter 201) for the state S including the servo state including the instruction, feedback, and the like including the positional deviation information of the servo control device 200 acquired by executing the machining program at the time of learning, based on the coefficients ω, τ, and σ.

Fig. 13 is a block diagram showing the machine learning unit.

In order to perform reinforcement learning as described above, as shown in fig. 13, the machine learning unit 505 includes a state information acquisition unit 551, a learning unit 552, a behavior information output unit 553, a cost function storage unit 554, and an optimization behavior information output unit 555.

The state information acquisition unit 551 acquires, from the servo control device 200, state information S including a command including positional deviation information of the servo control device 200 acquired by executing a machining program at the time of learning, feedback information, and other servo states, which are feedback information, based on coefficients ω, τ, and σ of the transfer function of the filter 201. The state information S corresponds to the environmental state S in Q learning.

The state information acquisition unit 551 outputs the acquired state information S to the learning unit 552.

In addition, initially, coefficients ω, τ, and σ at the time of starting Q learning are generated in advance by the user.

The learning unit 552 is a unit that learns the value Q (S, a) when a certain behavior a is selected in a certain environmental state S. Specifically, the learning unit 552 includes a report output unit 5521, a cost function update unit 5522, and a behavior information generation unit 5523.

The return output unit 5521 is a unit for calculating a return in the case where the action a is selected in a certain state S. Here, the positional deviation, which is a state variable in the state S, is denoted by e (S), and the positional deviation, which is a state variable related to the state information S 'changed from the state S according to the behavior information a (coefficients ω, τ, and σ of the transfer function of the filter 201), is denoted by e (S').

The evaluation function f uses, for example, the same evaluation function as that of equation 3 or equation 4 used in the operation evaluation unit 520.

At this time, when the value f (e (S)) of the evaluation function f when the servo control device 200 is operated based on the corrected filter 201 related to the state information S' corrected based on the behavior information a is larger than the value f (e (S)) of the evaluation function f when the servo control device 200 is operated based on the filter 201 related to the state information S before correction based on the behavior information a, the return output unit 5521 sets the value of return to a negative value.

On the other hand, when the value f (e (S)) of the evaluation function f when the servo control device 200 is operated based on the corrected filter 201 related to the state information S' corrected based on the behavior information a is smaller than the value f (e (S)) of the evaluation function f when the servo control device 200 is operated based on the filter 201 related to the state information S before correction based on the behavior information a, the return output unit 5521 sets the value of return to a positive value.

When the value f (e (S)) of the evaluation function f when the servo control device 200 is operated based on the corrected filter 201 related to the state information S' corrected based on the behavior information a and the value f (e (S)) of the evaluation function f when the servo control device 200 is operated based on the filter 201 related to the state information S before correction based on the behavior information a are equal, the return output unit 5521 sets the return value to 0.

In addition, when the value f (e (S)) of the evaluation function f in the state S' after the execution of the behavior a is larger than the value f (e (S)) of the evaluation function f in the previous state S, the negative value may be increased in accordance with the ratio. That is, the negative value may be increased in correspondence with the degree to which the value of the value f (e (S)) of the evaluation function f of the state S' is increased. Conversely, if the value f (e (S)) of the evaluation function f in the state S' after the execution of the action a is smaller than the value f (e (S)) of the evaluation function f in the previous state S, the positive value may be increased in accordance with the ratio. That is, the positive value may be increased in accordance with the degree to which the value of f (e (S)) is decreased.

The cost function updating unit 5522 performs Q learning based on the state S, the behavior a, the state S' when the behavior a is applied to the state S, and the reported value r calculated as described above, thereby updating the cost function Q stored in the cost function storage unit 554.

The updating of the cost function Q may be performed by either online learning or batch learning or by small batch learning.

Online learning is a learning method in which updating of the cost function Q is performed immediately every time the state S transitions to a new state S' by applying a certain behavior a to the current state S. In addition, batch learning is a learning method in which a certain behavior a is repeatedly applied to a current state S to shift the state S to a new state S', data for learning is collected, and the value function Q is updated by using all collected data for learning. Further, the batch learning is a learning method in which the cost function Q is updated every time learning data to a certain extent is accumulated in the middle of the online learning and the batch learning.

The behavior information generating unit 5523 selects the behavior a during Q learning for the current state S. During Q learning, the behavior information generating unit 5523 generates behavior information a for performing operations of the correction coefficients ω, τ, and σ (corresponding to the behavior a during Q learning), and outputs the generated behavior information a to the behavior information output unit 553. More specifically, the behavior information generating unit 5523 outputs, for example, behavior information a, which is added or subtracted incrementally to or from the coefficients ω, τ, and σ included in the state S, to the behavior information output unit 553. The behavior information a is correction information of the coefficients ω, τ, and σ.

When the increase or decrease of the coefficients ω, τ, and σ is applied and the positive return (return of the positive value) is returned, the behavior information generating unit 5523 may select, as the next behavior a ', the behavior a' in which the value of the positional deviation is smaller by adding or subtracting the coefficients ω, τ, and σ incrementally as in the previous behavior.

Conversely, when a negative return (return of a negative value) is returned, the behavior information generating unit 5523 may select, as the next behavior a ', a behavior a' in which the positional deviation of the coefficients ω, τ, and σ is smaller than the previous value by adding or subtracting the coefficients ω, τ, and σ incrementally, in contrast to the previous behavior.

The behavior information output unit 553 is a unit that transmits correction information, which is a coefficient of the behavior information a output from the learning unit 552, to the filter 201. As described above, the filter 201 performs the micro-correction of the current state S, that is, the coefficients ω, τ, and σ of the transfer function currently set, based on the behavior information, and thereby shifts to the next state S' (that is, the corrected coefficients ω, τ, and σ).

The cost function storage 554 is a storage device that stores the cost function Q. The cost function Q may be stored as a table (hereinafter referred to as a behavior cost table) for each state S and behavior a, for example. The cost function Q stored in the cost function storage 554 is updated by the cost function updating unit 5522.

The optimization behavior information output unit 555 generates behavior information a (hereinafter referred to as "optimization behavior information") for maximizing the value Q (S, a) of the filter 201 based on the value function Q updated by the Q learning by the value function update unit 5522.

The optimized behavior information includes information of coefficients ω, τ, and σ of the transfer function of the correction filter 201, similarly to the behavior information output by the behavior information output unit 553 during Q learning.

The servo control device 200 can be operated to correct the coefficients ω, τ, and σ based on the correction information of the coefficients, thereby reducing the value of the positional deviation.

By using the machine learning unit 550 as described above, the adjustment of the coefficients ω, τ, and σ of the transfer function of the filter 201 by the servo control device 200 can be simplified.

The functional blocks included in the numerical controller 100, the servo controller 200, and the controller 500 are described above.

In order to realize these functional blocks, the numerical controller 100, the servo controller 200, and the controller 500 are provided with an arithmetic processing device such as a CPU (central processing unit). The servo control device 200 further includes an auxiliary storage device such as an HDD (hard disk drive) that stores various control programs such as application software and an OS (operating system), and a main storage device such as a RAM (random access memory) that stores data temporarily required by the arithmetic processing device when executing the programs.

In the numerical controller 100, the servo controller 200, and the controller 500, the arithmetic processing unit reads the application software and the OS from the auxiliary storage device, expands the read application software and OS in the main storage device, and performs arithmetic processing based on the application software and the OS. In addition, various hardware included in each device is controlled based on the calculation result. Thereby, the functional module of the present embodiment is realized. That is, the present embodiment can be realized by cooperation of hardware and software.

Since the machine learning unit 550 has a large amount of computation by machine learning, for example, a GPU (graphics processing unit) can be installed in a personal computer, and the GPU is used in computation processing by machine learning by a technique called GPGPU (General-purpose computing by graphics processing unit: general-Purpose computing on Graphics Processing Units), thereby enabling high-speed processing. Further, in order to perform higher-speed processing, a computer cluster may be constructed using a plurality of such GPU-mounted computers, and parallel processing may be performed by a plurality of computers included in the computer cluster.

(Second embodiment)

The machine control system of the present embodiment has the same configuration as the machine control system 10 shown in fig. 1 and 2, except for the configuration of the servo control device.

Unlike the servo control device 200 shown in fig. 2, the servo control device 200A of the present embodiment is provided with a second differentiator 210, a speed feedforward processing section 211, and an adder 213 without a filter 201.

The second differentiator 210 outputs a value obtained by differentiating the position command value and multiplying the position command value by a constant β to the speed feedforward processing section 211. The velocity feedforward processing unit 211 performs velocity feedforward processing as shown by a transfer function H(s) shown in mathematical formula 6 (hereinafter, expressed as formula 6) on the output of the second differentiator 210, and outputs the processing result to the adder 213 as a feedforward term. The coefficients c _i、d_j (X.gtoreq.i, j.gtoreq.0, X is a natural number) of equation 6 are the coefficients of the transfer function H(s).

[ Formula 6]

The machine learning unit 550 executes a learning processing program, and uses the values of the coefficients c _i、d_j (i, j++0) of the transfer function H(s) of the position deviation learning speed feedforward processing unit 211 obtained from the subtractor 202 to reduce the position deviation. Specifically, as described in detail in the first embodiment, the machine learning unit 550 learns, as the state S, a servo state including the values of the coefficients c _i、d_j (i, j+.0) of the transfer function of the speed feedforward processing unit 211 in the servo control device 200, and an instruction and feedback of the positional deviation information and the positional instruction of the servo control device 200 obtained by executing the machining program at the time of learning, and learns to select, as the value Q of the action a, the adjustment of the value of the coefficient c _i、d_j of the transfer function of the speed feedforward processing unit 211 in the state S.

Here, the shape of the movement trajectory specified by the machining program at the time of learning is, for example, an octagon as shown in fig. 11. Based on the position P3 and the position P4 of the movement locus shown in fig. 11, the vibration at the time of changing the rotation speed in the linear control is evaluated, and the influence on the positional deviation is investigated, thereby learning the coefficient of the transfer function H(s).

Here, the machine learning unit 550 learns the value of the coefficient c _i、d_j (i, j+.0) of the transfer function H(s) of the speed feedforward processing unit 211, but may learn the value of the coefficient a _i、b_j (i, j+.0) of the transfer function G(s) of the position feedforward processing unit 206 instead of this, or may learn both the coefficient of the transfer function H(s) and the coefficient of the transfer function G(s). The coefficient c _i、d_j (i, j+.0) of the transfer function H(s) of the speed feedforward processing section 211 and the coefficient a _i、b_j (i, j+.0) of the transfer function G(s) of the position feedforward processing section 206 are control parameters for determining the operation characteristics of the driving section of the machine 400 by the servo control device serving as the motor control device.

It is desirable that, when both the coefficient of the transfer function H(s) and the coefficient of the transfer function G(s) are learned, the machine learning unit 550 performs learning of the coefficient of the transfer function of the speed feedforward processing unit 211 and learning of the coefficient of the transfer function of the position feedforward processing unit 206, respectively, and the coefficient of the transfer function of the speed feedforward processing unit 211 located further inside (inner loop) than the position feedforward processing unit 206 is learned before the coefficient of the transfer function of the position feedforward processing unit 206 is learned. Specifically, the coefficient of the transfer function of the position feedforward processing section 206 is fixed, and the optimum value of the coefficient of the transfer function of the speed feedforward processing section 211 is learned. Then, the machine learning unit 505 fixes the coefficient of the transfer function of the speed feedforward processing unit 211 to the optimum value obtained by learning, and learns the coefficient of the transfer function of the position feedforward processing unit 206.

Thus, learning related to optimization of the coefficient of the transfer function of the position feedforward processing unit 206 can be performed under the condition of the velocity feedforward term optimized by learning, and variation in positional deviation can be suppressed.

The machining program for checking the operation state in the present embodiment may use the machining program whose movement locus is octagonal as shown in fig. 11, but other shapes may be used as long as the shape of vibration at the time of changing the rotation speed in the linear control can be evaluated. The operation of the control device 500 in the present embodiment is the same as that described in the first embodiment.

The numerical controller, the servo controller, and each component included in the controller can be realized by hardware, software, or a combination thereof. The servo control method and the control method performed by cooperation of the respective components included in the servo control device and the control device may be realized by hardware, software, or a combination thereof. Here, the implementation by software means implementation by reading and executing a program by a computer.

Various types of non-transitory computer readable media (non-transitory computer readable medium) can be used to store programs and supply computers. Non-transitory computer readable media include various types of tangible recording media (tangible storage medium). Examples of the non-transitory computer readable medium include magnetic recording media (e.g., floppy disks, hard disk drives), magneto-optical recording media (e.g., optical disks), CD-ROMs (read-only memories), CD-R, CD-R/W, semiconductor memories (e.g., mask ROMs, PROMs (programmable ROMs), EPROMs (erasable PROMs), flash ROMs, RAMs (random access memories)). In addition, programs may also be supplied to the computer through various types of transitory computer readable media (transitory computer readable medium).

The above-described embodiments are suitable embodiments of the present invention, but the scope of the present invention is not limited to the above-described embodiments, and can be implemented in various modifications without departing from the scope of the main aspects of the present invention.

In the above embodiment, an example in which a servo motor is used as the motor and a servo controller is used as the motor controller has been described. However, the present invention is not limited to this, and for example, a stepping motor may be used as the motor, and a control device that does not perform servo control may be used as the motor control device.

In the above embodiment, the control device acquires the positional deviation, which is information indicating the operation characteristics of the driving unit of the machine, from the servo control device, but may be acquired from an acceleration sensor attached to the machine when the information is acceleration information of the driving unit of the machine. That is, information indicating the operation characteristics of the driving unit of the machine can be obtained from the servo control device or the machine.

In the above embodiment, the control command output from the numerical controller to the motor controller is described as a position command, but the control command is not limited to a position command, and may be, for example, a speed command.

In addition, the structure of the mechanical control system is the following structure in addition to the structure of fig. 1.

< Deformation example in which the control device is provided outside the CNC device >

Fig. 15 is a block diagram showing another configuration example of the machine control system. The machine control system 10A shown in fig. 15 is different from the machine control system 10 shown in fig. 1 in that: n (n is a natural number of 2 or more) control devices 500-1 to 500-n are connected to n CNC devices 20A-1 to 20A-n via a network 600. CNC devices 20A-1-20A-n are coupled to machines 400-1-400-n. The CNC devices 20A-1 to 20A-n each have the same structure as the CNC device 20 shown in FIG. 1. The control devices 500-1 to 500-n each have the same configuration as the control device 500 shown in fig. 1. The machines 400-1 to 400-n each have the same structure as the machine 400 shown in fig. 2.

The CNC device 20A-1 is in this case communicably connected to the control device 500-1 in a one-to-one group. The CNC devices 20A-2 to 20A-n and the control devices 500-2 to 500-n are also connected to the CNC device 20A-1 and the control device 500-1 in the same manner. In fig. 15, n groups of CNC devices 20A-1 to 20A-n and control devices 500-1 to 500-n are connected via a network 600, but for n groups of CNC devices 20A-1 to 20A-n and control devices 500-1 to 500-n, the CNC devices and control devices of the respective groups may be directly connected via connection interfaces. The n groups of the CNC devices 20A-1 to 20A-n and the control devices 500-1 to 500-n may be provided in the same plant, for example, in a plurality of groups, or may be provided in different plants, respectively.

The network 600 is, for example, a LAN (local area network), the internet, a public telephone network, or a combination thereof, which is built in a factory. The specific communication method, which is a wired connection or a wireless connection, etc. in the network 600 is not particularly limited.

< Degree of freedom of System configuration >

In the modification example described above, the CNC devices 20A-1 to 20A-n and the control devices 500-1 to 500-n are communicably connected in a one-to-one group, but for example, one control device may be communicably connected to a plurality of CNC devices via the network 600, and the operation state checking operation and the machine learning of each CNC device may be performed.

In this case, the functions of one control device may be appropriately distributed to a plurality of servers. In addition, each function of one control device may be realized by a virtual server function or the like on the cloud.

In the case where n control devices 500-1 to 500-n corresponding to n CNC devices 20A-1 to 20A-n of the same model, the same specification, or the same series are present, the operation status checking operation and/or the machine learning result of each control device 500-1 to 500-n may be shared. This makes it possible to construct a more appropriate model.

Claims

1. A control device is characterized by comprising:

A machine learning unit that performs machine learning of control parameters that determine operation characteristics of a drive unit of a machine that is a drive target of a motor, and sets the control parameters to a motor control device that controls the motor;

an operation state check operation instruction unit that outputs an instruction to drive an operation state check operation of the motor control device in order to obtain an operation characteristic of the drive unit;

An operation evaluation unit that acquires information indicating an operation characteristic of the driving unit from the motor control device or the machine that is driven in response to an instruction of the operation condition checking operation, calculates an evaluation value from an evaluation function using the information, and stores the evaluation value in a storage unit in association with time information at the time of executing the operation condition checking operation or the operation information of the motor; and

A deterioration estimating unit that estimates deterioration of an operation characteristic of a driving unit of the machine based on the evaluation value calculated by the operation evaluating unit when the operation condition checking operation is performed and the evaluation value stored in the storage unit,

Wherein the degradation estimating unit instructs the machine learning unit to learn the control parameter based on a result of estimating degradation of the operation characteristic of the driving unit of the machine.

2. The control device according to claim 1, wherein,

The degradation estimating unit instructs a report unit that notifies degradation of the operation characteristics of the driving unit of the machine based on the estimation result of the degradation of the operation characteristics of the driving unit of the machine.

3. Control device according to claim 1 or 2, characterized in that,

The operation state check operation instruction unit outputs an instruction of an operation state check operation when a predetermined signal is input or according to a predetermined schedule.

4. Control device according to claim 1 or 2, characterized in that,

The operation state check operation instruction unit transmits an instruction of the operation state check operation to a numerical controller that outputs a control instruction to the motor controller.

5. A CNC device, comprising:

the control device of claim 4;

a motor control device that controls the motor; and

And a numerical controller for outputting a control command to the motor controller based on the instruction of the operation condition check operation outputted from the controller.

6. A control method executed by a control device provided with: a storage unit; a machine learning unit for machine learning control parameters that determine operation characteristics of a drive unit of a machine to be driven by a motor, and for setting the control parameters to a motor control device that controls the motor,

The following steps are performed by the control device:

An operation evaluation step of acquiring information indicating an operation characteristic of the driving unit from the motor control device or the machine driven in response to an instruction of the operation condition checking operation, calculating an evaluation value from an evaluation function using the information, and storing the evaluation value in a storage unit in association with time information at the time of executing the operation condition checking operation or the operation information of the motor; and

A degradation estimating step of estimating degradation of an operation characteristic of a driving unit of the machine based on the evaluation value calculated in the operation evaluating step and the evaluation value stored in the storage unit when the operation condition checking operation is performed,

The degradation estimation step further includes:

and a machine learning instructing step of instructing machine learning of the control parameter to the machine learning unit based on a result of estimation of degradation of the operation characteristic of the driving unit of the machine estimated in the degradation estimating step.

7. The control method according to claim 6, wherein,

The degradation estimation step further includes:

And a reporting step of reporting the deterioration of the operation characteristic of the driving unit of the machine based on the result of the estimation of the deterioration of the operation characteristic of the driving unit of the machine in the deterioration estimating step.